🚀 Feature Roadmap — Functional Capabilities Beyond MVP

[AliHamaSmart]

🚀 Feature Roadmap — Functional Capabilities Beyond MVP

🎯 Goal

Track the next wave of functional features for Netra (multi-tenant face-recognition attendance) beyond the current MVP. The MVP covers: email/JWT auth, kiosk device-token attendance, face enrollment (consent/UU PDP), schedule-window validation, dedup, geofence flagging + location capture, daily status roster, reports/exports, soft-delete/trash, onboarding, webhooks, realtime (WebSocket).

This issue is the umbrella tracker; each phase can be split into its own issue when picked up.

---

🔴 Phase 1 — Core Reliability (do first)

• 1.1 Recognition accuracy: upgrade buffalo_s → buffalo_l (detector det_10g + ResNet50 w600k_r50), retune MATCH_THRESHOLD, add quality gate (min det score / min face size). Blocker: current same-person similarity ~0.05.
• 1.2 Manual attendance correction / override (admin) — fix missed scans, on-duty-outside, wrong matches.
• 1.3 Leave / permission management (Izin / Sakit / Cuti / Dinas Luar) — new status so "Belum Hadir" ≠ "Izin"; reflected in roster + reports.
• 1.4 Fix test isolation flakiness (shared async engine across files) + reportlab in image (rebuild) + ensure CI green.

🟠 Phase 2 — Attendance & Scheduling Depth

• 2.1 Per-group schedules — assign schedules to class/department/shift group (not one tenant default).
• 2.2 Holiday calendar UI (rules already support holidays).
• 2.3 Geofence config UI in Schedule form (set center lat/lng + radius_m) — backend flag already implemented.
• 2.4 Multi-session / per-subject attendance (session schedule type exists; needs UI + flow).
• 2.5 Kiosk offline mode — queue punches when offline, sync on reconnect.

🟡 Phase 3 — Notifications & Communication

• 3.1 Outbound notifications (WhatsApp / Email / SMS) to parents/guardians/HR on late/absent.
• 3.2 Notification preferences per tenant (channels, thresholds, quiet hours).
• 3.3 Daily/period digest to homeroom teacher / manager.

📊 Phase 4 — Reporting & Analytics

• 4.1 Analytics dashboard — attendance trend (line), status breakdown (donut), peak hours, % present per group.
• 4.2 Per-user history + attendance percentage (for report cards / performance).
• 4.3 Scheduled/automated reports emailed to admins.
• 4.4 Audit log viewer UI (table audit_logs is populated; UI is a placeholder).

🔌 Phase 5 — Integration (client data linkage)

• 5.1 API key management + webhook management UI (let client apps pull attendance; external_id is the join key).
• 5.2 Bulk import (CSV/Excel) of users + external_id (validate dropdown/flow).
• 5.3 Push-back sync of attendance results to client SIS/HRIS (scheduled).
• 5.4 SSO / OIDC staff login (Google Workspace / Microsoft) — SSO model scaffold exists.

🧠 Phase 6 — Biometric & Kiosk

• 6.1 Re-enrollment workflow (faces drift over time — glasses, growth).
• 6.2 Quality gate + multi-face rejection before embedding.
• 6.3 Self-attendance via phone (selfie + geofence, per-user device token) — no physical kiosk.

🔐 Phase 7 — Security & Compliance (UU PDP)

• 7.1 Consent management + right-to-erasure (revoke consent, delete biometrics on request).
• 7.2 Data retention / auto-purge of old embeddings & images (partial purge exists for soft-delete).
• 7.3 2FA for admin accounts.

🏢 Phase 8 — SaaS Platform (super admin)

• 8.1 Billing / subscription (per user/tenant quotas, plan limits).
• 8.2 Per-tenant usage metrics (scans, storage, active users).
• 8.3 White-label branding per tenant (tenant config space already exists).

---

✅ Recommended order

1. Phase 1.1 (recognition accuracy) — core product blocker; everything else sits on this.
2. Phase 1.2 + 1.3 (manual correction + leave/permission) — makes attendance data correct & complete (distinguish alpa vs izin).
3. Phase 3.1 (outbound notifications) — highest perceived value for schools.

📝 Notes

• Geolocation requires HTTPS in production (works on localhost dev).
• Recognition similarity issue is the current hard blocker — see Phase 1.1.
• Related UI/UX track: .github/ISSUES/ui-ux-redesign-10-10.md.